Scientific Machine Learning

Comments · 280 Views

Traditional ML courses and the projects you do as part of it are not meant for engineers.

Traditional ML courses and the projects you do as part of it are not meant for engineers.

As a mechanical engineer or a physicist, why would I do a project on movie review analyzer or housing price prediction?

I would love to do a project which teaches me how to use ML to model fluid mechanics or black hole dynamics.

I want a field which combines ML with my domain knowledge.

Scientific ML is exactly that field.

I feel Scientific ML is one of the coolest techniques of the last 4-5 years.

There are 3 main pillars of Scientific ML:

(1) Neural ODEs

(2) Physics Informed Neural Networks (PINNs)

(3) Universal Differential Equations

It helped me transition from mechanical engineering to machine learning, and obtain a PhD at MIT in Machine Learning.

Any thoughts on Scientific ML or PINNs or Neural ODEs?

Comments